Author Topic: Repair worklog : Old rusty HP 3458A (Read 254756 times)

Dr. Frank · « **Reply #500 on:** January 06, 2016, 10:05:31 pm »

I think, that I've shown this stability measurement already.

3458A NPLC100, over 35h vs. an LTZ1000, 25k samples.
ppm of mean LTZ1000 value, i.e. related to 7.162..V
No ACAL in between, but small temperature variation only and no EMC disturbance in our basement.

A fine 3458A and also a LTZ1000 reference are well capable of < 0.2ppm drift stability on that timescale, and < 0.04ppm of noise.

A resolution of 0.1°C is sufficient for that.

Frank

TiN · « **Reply #501 on:** January 07, 2016, 04:18:39 am »

Hi Alan,

No, issue I have is not temperature drift. I spent days to make sure of it. And I'm very confident it's not drift of A9 LTZ reference output or something on A1 DC PCBA either.
I did left it drifting for few days before, and it had constant rate, going over 20ppm off, which is over any possible spec. K2001 connected to same source was showing stability in ppm level.

Anyhow, ManateeMafia had multiple 3458A's sampling 10V, and I finally got ahold to tidy-up and make page with summary graphs - 4 x HP3458A test of Fluke 732B. It's permalink https://xdevs.com/datalog_comp/, updated in realtime every 5 minutes, as new data comes in.

There are two caveats before you can study data graphs:

1. Sometimes data CSV's had delays, but graph shows as interpolated line between. For example there is no data at December 31, but line is there.
2. Time/date scale axis is not same on all graphs (as different meters were added into measurement at different days)

I don't know JavaScript or D3.js to figure how to fix these issues at this moment, so you warned

Lessons I've learn't so far from data:

Meter 04188 have drift +0.066ppm/day and tempco +0.066ppm/K
Meter 08451 (with real 002 option) have drift +0.145ppm/day and tempco -0.266ppm/K
Meter 01527 have drift or tempco 4ppm/day?. Not enough data to be sure (only 2 days history so far)
Meter 02157 with repaired A3 ADC have drift -0.05ppm/day and tempco -0.266ppm/K
Meter 02157 with good A3 ADC have very low drift (not much data to say for sure). Note finer ppm scale

I think this is very good data to answer question how good stability 3458A should look like...
Humidity, pressure and ambient temperature are also logged by standalone sensor hooked to same Pi, so you can nit-pick if there is any relation...

Dr. Frank · « **Reply #502 on:** January 07, 2016, 07:05:35 am »

TiN,
are all these measurements over several days done without ACAL DCV in between?

Frank

TiN · « **Reply #503 on:** January 07, 2016, 07:11:35 am »

That's correct, no ACAL anywhere.

P.S. Caveat with date scales fixed, now all X-axis on all graphs uniform!
P.P.S. Patched my nginx configs so CSV-files are served GZIPped now, page should load 2-3 times quicker (It's about 15MB of text data).

alanambrose · « **Reply #504 on:** January 09, 2016, 12:02:39 pm »

Hi Guys,

Apologies have been busy on something else

DrF - If you use a big enough averaging interval you will, of course, remove all noise in the limit - isn't this is just the hardware/software equivalent of a LPF. As you mentioned before, you might have to measure noise in a way that doesn't pick up drift also. I chose 2 nplc because I wanted to pick up the DC-10Hz noise. Re temp resolution at 0.1C - for any reading you can always scrape information out of the 'quantisation noise' but it would be helpful if the unnecessary noise wasn't there in the first place.

MM - That's a pretty interesting test - are you planning to run the correlations and try and unpick the temperature / time / pressure / humidity coefficients? These are all relative to the Fluke i.e. assuming the Fluke is fully stable?

TiN - Re 20ppm - I didn't realise the drift on the 'evil' meter was quite that high. In a sense - as it is so high it should make it easier to find. Did you eyeball the temp/drift coefficients for the 4x3458As or run some regressions?

Did anyone figure out how/where the 3458A internal temperature is measured? I'm still thinking some controlled tests on tempco would be interesting i.e. making deliberate changes in ambient and getting an accurate feel for the tempco so it can be taken out of the time drift calc. I see the spec suggests 0.5ppm per day and also 0.5ppm per C for 10V which suggests MMs meters are well within spec - except maybe for #01527 and maybe it will also turn out good.

A.

TiN · « **Reply #505 on:** January 09, 2016, 12:21:52 pm »

Temp sensor is plain LM35DZ in TO92 located on front left side of A1 PCB, routed thru digital mux and K6 relay to ADC.

alanambrose · « **Reply #506 on:** January 09, 2016, 12:33:10 pm »

Thanks, I see - U21. You have probably figured out what this is too

TiN · « **Reply #507 on:** January 09, 2016, 02:34:52 pm »

Track&Hold capacitor for ACV. It's shown on schematics as well.

alanambrose · « **Reply #508 on:** January 10, 2016, 06:42:24 pm »

Ah yes, not a delay line or an inductor as it only has one end nailed down

For #08451 I'm getting these rough results as individual simple linear regressions. For the second set, I took the time drift out and tried the other factors again. Results (for the simple linear model) suggest no significance on anything except time drift.

BTW regressing internal temp vs. external temp gets:

internal = external * 66% + 19.4C (with R squared = 65%)

I guess that says something about the 3458A temperature regulation characteristics.

A.

Results are here: http://anagram.net/nuts/MeterDrift/10v_3458_nplc200_mm_08451_opt002.htm

Analysis is here: http://anagram.net/nuts/MeterDrift/10v_3458_nplc200_mm_08451_opt002.xlsx

Dr. Frank · « **Reply #509 on:** January 10, 2016, 07:31:22 pm »

Quote from: alanambrose on January 09, 2016, 12:02:39 pm

Hi Guys,

..
DrF - If you use a big enough averaging interval you will, of course, remove all noise in the limit - isn't this is just the hardware/software equivalent of a LPF. As you mentioned before, you might have to measure noise in a way that doesn't pick up drift also. I chose 2 nplc because I wanted to pick up the DC-10Hz noise. Re temp resolution at 0.1C - for any reading you can always scrape information out of the 'quantisation noise' but it would be helpful if the unnecessary noise wasn't there in the first place.

...A.

It was not the noise that was in question here.
The noise parameter of the 3458A is covered by TiNs - zero input - noise measurement project, about the LTZ100s noise would be interesting also.

No, the question HERE was about timely stability, and I think I have answered that exactly and straight forward

Frank

TiN · « **Reply #510 on:** January 13, 2016, 01:19:40 am »

Next step...

So far it's good. Let's see if it stays that way next days.

ManateeMafia · « **Reply #511 on:** January 13, 2016, 02:08:08 am »

The phoenix has risen, again. Looking forward to seeing some good readings. I don't suppose you have a way to monitor the environment?

TiN · « **Reply #512 on:** January 13, 2016, 04:20:24 am »

I need to overclock my hours to 48 a day at least. Will hook sensor to Pi tonight to add PTH data as well.

So far tempco non-measureable, and reading stay within 0.1ppm window over last 8 hours (averaged).
Interesting thing, I added my data into overall graphs, and it's easy to see 3245A's low-frequency oscillation, compared to way more stable 732B output.

Time scale is matched.
It's time to make LTZ-based 10V source then.

alanambrose · « **Reply #513 on:** January 14, 2016, 02:28:36 pm »

Well congrats, nice job

The 3245A wobble is strange - I added a moving average to my 3245/3458 drift/noise measure above and updated the graph in the post to try and show any similar wobble (Excel is dumb and puts on a trailing moving av. instead of a centered one.) I can't see one.

Any final theroies what was wrong with the old A3?

A.

TiN · « **Reply #514 on:** January 14, 2016, 02:52:29 pm »

That graph already with averaging.
I think it's something in lab coupling to meter. We talking fractions of ppm now. Life is not easy in that range...

No theories, old A3 have U180 hybrid asic drifting.
I have few ideas for it, but will do it later.

I hope to get my VISA fixed on PC tomorrow, so I can finally run your app for 3245+3458A. Really want to compare your results with mine.

alanambrose · « **Reply #515 on:** January 16, 2016, 04:03:19 pm »

Don't suppose the wobble could be linked to mains frequency? In the meantime FYI:

That's with the source well warmed up beforehand. So 1.5h for 1ppm of eventual reading. For the last couple of hours of that test the tempco was +0.4ppm / C.

A.

alanambrose · « **Reply #516 on:** January 18, 2016, 09:47:20 pm »

And apropos of nothing (except it's more drift data) data for an Advantest R6144. Drift is apparently:

+3ppm per day
+0.4ppm / C
+0.26ppm / Hz

Total R-squared is 96%.

Analysis is here:

http://anagram.net/nuts/R6144/HP3458A%20driftandnoise%202016-01-15%2018-19-43.xlsx

I don't remember enough regression stuff to figure out how relevant the last two variables are.

TiN · « **Reply #517 on:** January 19, 2016, 04:17:09 am »

Hi Alan,

Can you also measure 3245 set at +2.5VDC over 4-5 hours?
I saw some strange stuff, like drift up ~0.5ppm/hour. EDC MV106 2.5VDC was stable, so it's was not 3458.

Also seems like LTZ1000CH I have now in A9 still have shot pink noise, ~0.3ppm jumps. I'd expect all zeners have this popcorn noise, but some smaller than others.
I don't really want binning LTZ craze, so will look into other possible solutions, like zener current change maybe.
Need add probe to measure LTZ output directly, to be sure though.

alanambrose · « **Reply #518 on:** January 19, 2016, 12:02:24 pm »

Sure will do.

A.

alanambrose · « **Reply #519 on:** January 23, 2016, 05:42:16 pm »

>>> Can you also measure 3245 set at +2.5VDC over 4-5 hours?

OK (this is all ppm of 2.5V):

+0.8ppm per day
+1.1ppm / C
-0.41ppm / Hz (note minus sign)

Total R-squared is 90%. Noticeably more noise than R6144 (0.89ppm compared with 0.13) - although the R6144 was at 10V. A couple of things occurred to me:

+ not sure why you chose 2.5V but I guess close to ref voltage we're getting noise of ref plus dac, amplifier etc, close to 0V we're getting noise of amplifier mostly, some dac, no ref?

+ apologies I don't remember in which of these threads and exactly who suggested vlf fft - but talking to a mate here today - that sounds interesting as drift could be partly thought of as vlf 1/f noise, so a fft characterisation might be interesting?

>>> I saw some strange stuff, like drift up ~0.5ppm/hour.

Well part of the graph below shows e.g. drift of ~0.8ppm over 2h (from 10am 'til noon) so depending on the tempco of the EDC and the temp change over the measuring period, maybe that's possible? What ref does the EDC use?

Analysis is here:

http://anagram.net/nuts/3245A/HP3245A driftandnoise 2016-01-20 21-06-24.xlsx

TiN · « **Reply #520 on:** January 23, 2016, 07:10:55 pm »

Thanks, I see no drift on your with constant temperature, that's what I needed.
EDC using selected zener diode (1N829 or similar, TC-compensated), and it showed ~1.1ppm/C before, but question is not about it, it's about 3245.

2.5V output was due to my current experiment needs, since I was trying to use 3245A as stable DC source, but discovered that reading going up and up.
on 10V it's stable as this data suggests

TiN · « **Reply #521 on:** March 07, 2016, 05:37:13 pm »

Everyone forget about this thread already, eh?

I had the machine running 24/7 last 2 months, and last few weeks temps were going higher and higher. Yesterday was really sunny day, ambient in the room reached +30°C, and meter TEMP? output peaked at +46C. Yikes, that's bit too hot, especially due to my A9 reference running with lowered temp.

I moved meter other shelf, and in process of doing so, checked fan filter.. I just cleaned it in january.
Photos better than any words:

Moral of the story? Keep you 8.5-digit DMMs clean, or all your ppm's would run away....

pelule · « **Reply #522 on:** March 07, 2016, 07:52:23 pm »

That's what confuses me at the most HP/Agilent tools. The fan sucks outside air into the tool (inclding all potential dirt). The filter is nice, but not protective at that size of the mash.
Anybode ever thought of changing the FAN direction to blow out the air?
BR
PeLuLe

PA0PBZ · « **Reply #523 on:** March 07, 2016, 08:24:50 pm »

Quote from: pelule on March 07, 2016, 07:52:23 pm

Anybode ever thought of changing the FAN direction to blow out the air?

So the dirt will enter all the other (unfiltered) openings in the case?

plesa · « **Reply #524 on:** March 07, 2016, 08:33:25 pm »

What about actively cooling the top and bottom covers with water pipeline and use fan for internal air circulation only?


EEVblog Main Site	EEVblog on Youtube	EEVblog on Twitter	EEVblog on Facebook	EEVblog on Odysee

Author Topic: Repair worklog : Old rusty HP 3458A (Read 254756 times)

Share me